Josh Willis replies to comment on LQG and the diffeomorphism group

In summary, there has been a lot of internet discussion about the paper "Quantum gravity, shadow states, and quantum mechanics" by Ashtekar, Fairhurst, and Willis (gr-qc/0207106). One of the authors, Josh Willis, responded on the Physics Forums today in a thread about loop quantum gravity and the diffeomorphism group. The discussion revolves around the relationship between group averaging and anomalies in LQG and the different approaches of Rovelli and Thiemann. Willis explains their approach using non-standard unitary representations of the Weyl-Heisenberg algebra and how it relates to LQG. Urs Schreiber, who has been involved in the discussion, offers a summary of
  • #1
marcus
Science Advisor
Gold Member
Dearly Missed
24,775
792
There has been a lot of internet discussion of a paper by Ashtekar, Fairhurst, and Willis called "Quantum gravity, shadow states, and quantum mechanics" gr-qc/0207106, and one of the authors, Josh Willis, replied on SPR today (in the thread "LQG and diffeomorphism group cocycles")

Here is a representative exerpt of the longer post:

--------exerpt from Willis post------

Well, since Urs keeps referring to this paper I co-authored, I feel I
ought to respond with what I see.

So I will start by replying to the remarks about gr-qc/0207106, and
then work backwards towards some more general remarks about the
relationship between group averaging and anomalies, as I see it. This means I take the quoted article in reverse order. My apologies if this is confusing, but I think in the long run it will be clearer.

So, let me start with:

Urs Schreiber wrote:
> Sometimes LQG papers like gr-qc/0207106 are mentioned which supposedly show
> that the LQG-like 'form of quantization' reproduces ordinary quantization in
> some limit. But having a closer look at this paper shows that this _only_
> works when the usual quantum corrections are copied to the LQG-formalism
> (lower half of p.14). This is however not the case in the 'LQG-string' or
> the LQG treatment of the spatial diffeo constraints of 3+1d gravity.

I do not understand your assertion that the "quantum
corrections" are copied into the model system we study, and that this
is somehow very different from what is done with, for instance, the
diffeomorphism constraint in LQG.

So, for people who haven't read page 14 of gr-qc/0207106, let me
summarize quickly what is going on.

And let me start by saying that page 14 of that paper is the wrong
place to be looking to begin with: there we are finding candidate
semiclassical states, not considering commutation relations. Instead,
look at pages 7 and 8, to start.

We are looking at the one-dimensional point particle in quantum
mechanics. We have not yet considered any particular Hamiltonian, but
are instead at this point in the paper looking only at the canonical
commutation relations.

One normally learns these as:

{q,p} = 1

and then seeks in the quantum theory for self-adjoint operators
\hat{q} and \hat{p} that satisfy the corresponding relations:

[\hat{q},\hat{p}] = i\hbar

But these commutation relations among the p's and q's of course imply
commutation relations among the exponentiated operators, which---if
\hat{q} and \hat{p} are self-adjoint---will be unitary. This is the
relation:

(*) U(l)V(m) = exp{-ilm}V(m)U(l)

where:

U(l) = exp{il\hat{q}}
V(m) = exp{im\hat{p}}

The relation (*) is that of the Weyl-Heisenberg algebra, and what we
do in our paper is look at a non-standard unitary representations of
this algebra, rather than looking at a self-adjoint representation of
the CCR. The particular representation we look at is motivated by
analogy with LQG, and the main point of the paper is to look at the
low-energy limit of this theory, using techniques that it is hoped
will be relevant for LQG. However, as that does not seem to be what
the questions are about at the moment, I will not say more about that
here.

And this is the key point: the relation (*)---including the factor of
exp{-ilm}, which is what in other places on the net Urs seems to think
is put in by hand, i.e. by "knowing" what the quantum theory should
be---is in fact dictated *classically* by the Poisson algebra of the
basic observables. Any representation by unitary operators which
purports to be a canonical quantization *must* have this relationship,
and in particular the representation we consider does (as of course
also does the standard Schroedinger representation)...

--------end quote-------
 
Last edited:
Physics news on Phys.org
  • #2
It is fascinating that this discussion should relate in one way or another the spatial diffeomorphism group in LQG, and how LQG implements the diff-invariance found in General Relativity.
Especially interesting, I think, since Rovelli's treatment does not use the diffeomorphism group but rather the "almost smooth" group Diff*.
One result of extending the idea of diff-invariance is "separability"---making the kinetic state space of LQG have a countable basis.
A new paper of Rovelli and Fairbairn underscores this shift in the way spatial diff-invariance is handled and it seems to have interesting mathematical consequences ("Separable Hilbert space in LQG" gr-qc/0403047).

This is an area where LQG is in flux and where there are deep differences between Rovelli's and for example Thiemann's approaches, some of which is described in section 4.2 of the Rovelli/Fairbairn paper. It has become difficult to say frankly just what a "LQG-like" handling of the diffeomorphism constraint should be, or if there is one unique way.

Several people have joined the discussion at SPR, and indeed there has been a lot of talk about this same thing here at PF too, not to mention other internet forums.
 
Last edited:
  • #3
I like his characterization of group averaging, which you may remember we all made some heavy weather over. It STARTS with unitary products. We finally got there, and it's nice to know we weren't wrong.
 
  • #4
Urs reply today on SPR (brief sample)

Unfortunately, since these SPR posts are long, I can only copy representative samples. Later I will try to include links to the Cornell SPR archive. I copied about a third of Josh previous post and will show an even smaller fraction of Urs here, just to give a sample.

Today Urs replied with noteworthy civility and clarity. It raises hopes in me that this could be a very helpful discussion to listen to.
Here is a brief sample of today's post by Urs on SPR:

----quote---
...I do understand why the Weyl algebra (equation III.1) in your paper looksthe way it does, including that factor. I understand that this factor is not put in by hand. My point is that another factor is.

> How does one see this? By (as Urs has mentioned elsewhere) the
> Campbell-Baker-Hausdorff theorem, which let's me calculate what (*)
> should be, entirely in terms of iterated commutators of p's and q's.
> For the CCR, this is in fact the easiest way to go, because the CBH
> series stops after the first commutator.

I understand this and totally agree. But I think for those following this one should emphasize that the BCH theorem applied to exp(p)exp(q) can only serve to motivate certain steps in the LQG construction, because the p's and q's do not exists (not both at least) in your framework.

> But for more complicated Lie algebras, as for example the
> diffeomorphism algebra in gravity, the CBH series does not terminate
> and this is not a terribly enlightening way to view the problem.
>
> But it's also not necessary, because a self-adjoint representation of
> a Lie algebra exponentiates to a unitary representation of the Lie
> group. So what one should check instead is that one has an honest
> group representation, that is, that:
>
> U(g_1)U(g_2) = U(g_1 * g_2)
>
> And this *does* hold for the diffeo group in LQG.

Yes, as it does for Diff(S^1) x Diff(S^1) in Thomas Thiemann's 'LQG-sting'. By construction. I.e. operators U can be found which do satisfy this relation. I do understand and agree that these operators can be constructed.

The problem is, that these operators don't capture the usual quantum
effects. They do not come from what ordinarily is called 'canonical
quantization'. That's because the ordinary prescription of 'canonical
quantization' at some point says that we are to promote the classical
canonical coordinates and momenta to self-adjoint operators on some Hilbert space. This step is explicitly violated in LQG, since half of these objects are not represented as operators at all.

So with this step violated, something else has to be found to replace it. Among other things, I claim that this replacement introduces a large ambiguity.

Namely whenever there is now an object in the classical theory in the
enveloping algebra of p and q, you don't know precisely how to quantize it (even modulo operator ordering issues) since one of p and q is not represented on the Hilbert space.

I think that this is a problem both in gr-qc/0207106 as well as in the
'LQG-string'. Here is why:...

----end quote, read the rest on SPR---
 
  • #5
Originally posted by selfAdjoint
I like his characterization of group averaging, which you may remember we all made some heavy weather over. It STARTS with unitary products. We finally got there, and it's nice to know we weren't wrong.

It's great to have some of the same topics revisited in a slightly different light. I am hoping for a better understanding this time thru. Must say I do like Willis style, calm, and the care with which he organizes and explains.
 
  • #6


Originally posted by marcus
Unfortunately, since these SPR posts are long, I can only copy representative samples.

This is nonsense and you'll forgive me if I don't trust your judgement of what a "representative sample" would be. Here's the entire post:

"Josh Willis" <jwillis@gravity.psu.edu> schrieb I am Newsbeitrag
news:c2slkk$g2k$1@f04n12.cac.psu.edu...
> Well, since Urs keeps referring to this paper I co-authored, I feel I
> ought to respond with what I see.

Many thanks for your answer!

> I do not understand your assertion that the "quantum
> corrections" are copied into the model system we study, and that this
> is somehow very different from what is done with, for instance, the
> diffeomorphism constraint in LQG.

I'll try tomake my concern more precise below.

> (*) U(l)V(m) = e^{-ilm}V(m)U(l)

[...]

> The relation (*) is that of the Weyl-Heisenberg algebra, and what we
> do in our paper is look at a non-standard unitary representations of
> this algebra, rather than looking at a self-adjoint representation of
> the CCR.

Ok.

> And this is the key point: the relation (*)---including the factor of
> e^{-ilm}, which is what in other places on the net Urs seems to think
> is put in by hand,

No, sorry, that's a misunderstanding of what I was saying. I am grateful
that you took the time to answer so that I can try to clarify this.

I do understand why the Weyl algebra (equation III.1) in your paper looks
the way it does, including that factor. I understand that this factor is not
put in by hand. My point is that another factor is.

> How does one see this? By (as Urs has mentioned elsewhere) the
> Campbell-Baker-Hausdorff theorem, which let's me calculate what (*)
> should be, entirely in terms of iterated commutators of p's and q's.
> For the CCR, this is in fact the easiest way to go, because the CBH
> series stops after the first commutator.

I understand this and totally agree. But I think for those following this
one should emphasize that the BCH theorem applied to exp(p)exp(q) can only
serve to motivate certain steps in the LQG construction, because the p's and
q's do not exists (not both at least) in your framework.

> But for more complicated Lie algebras, as for example the
> diffeomorphism algebra in gravity, the CBH series does not terminate
> and this is not a terribly enlightening way to view the problem.
>
> But it's also not necessary, because a self-adjoint representation of
> a Lie algebra exponentiates to a unitary representation of the Lie
> group. So what one should check instead is that one has an honest
> group representation, that is, that:
>
> U(g_1)U(g_2) = U(g_1 * g_2)
>
> And this *does* hold for the diffeo group in LQG.

Yes, as it does for Diff(S^1) x Diff(S^1) in Thomas Thiemann's 'LQG-sting'.
By construction. I.e. operators U can be found which do satisfy this
relation. I do understand and agree that these operators can be constructed.

The problem is, that these operators don't capture the usual quantum
effects. They do not come from what ordinarily is called 'canonical
quantization'. That's because the ordinary prescription of 'canonical
quantization' at some point says that we are to promote the classical
canonical coordinates and momenta to self-adjoint operators on some Hilbert
space. This step is explicitly violated in LQG, since half of these objects
are not represented as operators at all.

So with this step violated, something else has to be found to replace it.
Among other things, I claim that this replacement introduces a large
ambiguity.

Namely whenever there is now an object in the classical theory in the
enveloping algebra of p and q, you don't know precisely how to quantize it
(even modulo operator ordering issues) since one of p and q is not
represented on the Hilbert space.

I think that this is a problem both in gr-qc/0207106 as well as in the
'LQG-string'. Here is why:

In the text leading to equation (IV.5) of gr-qc/0207106 you are looking for
a way to express the notion of 'coherent state' in the LQG-like language,
where you characterize a coherent state by its property of being an
eigenstate of the annihilation operator

a ~ x + ip

(which is one of several possibilities of characterizing coherent states).

This is an object in the algebra generated by p and q. It happens to have no
operator ordering problems so the usual quantization of this guy is obvious.

But now in your framework p is not representable, so something else has to
be done. Some object in the algebra of operators that you do have must be
identified as being the 'quantization' of the annihilation operator.

Due to the Weyl-algebra context, what you really want is the exponentiation
of the adjoint of the annihilation operator, i.e. an analog of

exp(a^dag) .

This has to be expressed somehow using the operators x and V that do exist
on your Hilbert space. You do this by looking at the usual quantization of
this object and read off the expression above equation (IV.5) from this,
e.g.

exp(a^dag) -> exp(x)V(-1/2)e^(-1/4) .

But why don't you use for instance

exp(a^dag) -> exp(x)V(-1/2)

or

exp(a^dag) -> exp(x)V(-1/2)e^(42) ?

This is the factor which I am talking about and which is put in by hand. It
is _not_ put in by hand in the _usual_ quantization, because there the
e^(-1/4) comes from the BCH theorem. But in your rep there is no p and q
which could enter the exponent in the BCH theorem. You are just copying this
plausible looking result to your representation.

And from reading the paper I got the impression that you would agree that
here you are fixing an ambiguity by hand. Because on p.14 it says:

"Collecting these ideas motivated by results in the Schroedinger
representation, we are now led to seek the analog...".

I do agree that this is indeed the most natural choice for this simple
system. My problem is that in more difficult cases, where there is mayber no
working Schroedinger representation from which ideas can be motivated, how
do you fix such ambiguities then?

I noted that in the 'LQG-string' as well as for the diffeo-constraints of
3+1d LQG gravity the analog of this ambiguity is fixed by using ideas
motivated by results not of the Schroedinger rep but of the classical
Poisson algabra.

Here is why:

When quantizing the Nambu-Goto action by LQG-like methods we are faced with
a problem very similar to the representation problem of the annihilation
operator "a" in the above example. Namely, there we have the classical
Virasoro constraints L_m, which are objects bilinear in the canonical p and
q.

In the standard quantization these p and q are promoted to operators and no
matter how you deal with the operator ordering ambiguity the resulting
quantum versions of the L_m feature the anomaly, which is a commutator
effect very much like the factor discussed above.

But now in Thomas Thiemann's 'LQG-string' quantization, as in your paper
above, the p and q are not represented as operators. Hence he has to look
for an _analog_ of the usual quantization, much like you construct an analog
of exp(a^dag), as discussed above.

So Thomas Thiemann considers the classically exponentiated Virasoro
generators and models his quantum operators U(phi) on these. He _could_ have
proceeded along the lines of your paper and "collected ideas motivated by
the Schroedinger representation to seek an analog of" the exponentiated
Virasoro generators in the usual quantization instead. This way he would
have had the anomaly (in terms of the group instead of the algebra, but
still).

But he chooses not to. He instead collects ideas motivated by the classical
action of the Virasoro generators. And that's how the anomaly disappears.

I'd very much enjoy hearing your opinion on these assessments. If I am
wrong, I'd very much want to be corrected. Thanks.
 
  • #7
I very much appreciate it that Josh Willis has posted a reply. Many thanks to him!

I haven't got an answer from A. Ashtekar so far, which I had asked by private email about his opinion on the LQG string. But he will give a talk here at the DPG Symposium tomorrow here in Ulm. With a little luck I find a chance to speak to him during this weak. Let's see.
 
  • #8
Originally posted by Urs
I very much appreciate it that Josh Willis has posted a reply. Many thanks to him!
...
I am happily looking forward to our hearing more from Urs about the
Spring Conference of the German Physical Society in Ulm----where he must be at this moment. Here is an exerpt from Urs post on CT:

---quote---
I am on my way to the spring conference of the German Physical Society...

Frühjahrstagung der Deutschen Physikalischen Gesellschaft

...
...

Ok, who else will bet there? There are many LQG people. A. Ashtekar will give a general talk on LQG for non-specialist. Bojowald of course will talk about what is called ‘Loop Quantum Cosmology’. With a little luck I find an LQGist willing to discuss the ‘LQG-string’ with me.

I would also like to talk to K.-H. Reheren, who has announced a talk on algebraic boundary CFT, about Pohlmeyer invariants, but I am not sure if he considers it worthwhile talking to me… :-/
----end quote---
 
  • #9
The latest news from my quest to understand LQG can be found here.
 
  • #10
Urs report, part 1

Originally posted by Urs
The latest news from my quest to understand LQG can be found here.

Urs report from the DPG Spring Conference is exciting and I copy here his reports from Sunday and Monday sessions:
--------quote from Urs CT post---------
Posted by: Urs Schreiber on March 14, 2004
Re: Sunday

Heard a very interesting and even moving, while unostentatious, talk by C. Yang about Albert Einstein, his ideas and his life.
Encountering Yang reminded me of the joke

Q: ‘Who invented the sledge hammer?’

A: ‘Mr. Sledge.’

It is almost like meeting somebody called ‘car’ or ‘lightbulb’, if you know what I mean. It’s kind of amazing.

Before the talk I met Rüdiger Vaas, who is science journalist for the german popular science journal Bild der Wissenschaft and is working a lot on reporting about research in quantum gravity. I had first met him at the Strings meet Loops symposium last year. His report on ‘Strings meet loops’ will appear in the next issue of BdW.

He tells me that more than half of the 12 pages long article will be concerned with LQG and in particular with M. Bojowald and his ‘Loop Quantum Cosmology’. Considering that also the recent issue of Scientific American had an article by Lee Smolin on LQG, which of course can be found translated in ‘Spektrum der Wissenschaft’, and considering that Spektrum and BdW are the two leading german journals for popular science, this gives an impressive amount of public attention for LQG here in Germany. Maybe there is a general tendency. The DPG Symposium here in Ulm is clearly dominated by LQG contributions. Kind of amazing when one is involved in the current discussion about the conceptual viability of LQG.

I mean, ok, it is not established that string theory will survive experimental tests and everybody is free to believe that it will not. But at least it is clearly about theoretical physics. All kinds of concepts in string theory will definitely survive in and enrichen theoretical physics even if it might turn out that gravitons are not excitations of some string.

But currently I am not so sure that LQG is even theoretically about physics.

But of course Bojowald’s claim to be able to connect quantum gravity with experimental MBR data is enchanting.

It would be great if for instance string cosmology could come up with a similarly nice cosmological model which removes and clarifies the initial singularity. Most of what I have heard so far about string cosmology was pretty disappointing. Veneziano’s pre-big bang model and similar scenarios always reminded me of the ‘then a miracle occurs’ mechanism. The interesting transition of the two classical branches remains a mystery.

Maybe Jacques Distler can increase my faith in string cosmology by reporting interesting results from the conference Cosmology and Strings that he mentions in his latest musings.
---to be continued---
 
Last edited:
  • #11
Urs report part 2

Urs report is spritely and entertaining. One sees what the purpose of these conferences is (for the grad student----perhaps the most essential perspective)
I quote in full:

-----Urs report for Monday----
Monday
This has been a very intensive day.

When I arrived (late) in A. Ahstekar’s talk this morning I had only five hours of sleep behind me, but I made it to the first coffee break without major casualties and resupplied myself with caffeine.

Ashtekar gave a general introductory lecture on LQG. Afterwards Peres talked about ‘Quantum Information and Relativity Theory’, being concerned with problems such as a quantum measurement in one frame may come before the measured event in another frame.

After the coffee break Clifford Will gave an extremely enjoyable talk on experimental tests of gravity. As I have said here, I learned that LISA will not be able to see primordial gravitational wave backgrounds due to the noise made by binary star systems in our galaxy.

After lunch the parallel sessions started. I had a problem, because I wanted to listen to the NCG stuff which was parallel to a session in which Folkert Müller-Hoissen gave a talk. Now, Eric and I have spent a lot of time with extending and generalizing the work by Dimakis and Müller-Hoissen and I had never met these authors before, so I decided to ditch the NCG session and sit in on ‘Symmetries, Integrability and Quantization’. It was sort of interesting, though I later learned that this way I missed a talk about NCG on Lorentzian spacetimes, which I really regret to have missed. I’ll need to talk to those people in private tomorrow.

Anyway, my hope was to get hold of Müller-Hoissen after the talks and get on his nerves by talking about discrete differential geometry. But unfortunately he was inolved in a discussion with somebody else about something else, which went on and on and on…

Finally I decided not to wait any longer and ran to the lecture auditorium to catch at least the second half of C. Fleischhack’s talk on ‘Progress and Pitfallls of LQG’. This was a very technical talk with lots of formulas with a huge amount of indices and symbol decorations. I have made a photograph of the point where he puts on the transparancy which says that now the spatial diffeomorphism constraints are solved. I believe that the field of LQG would maybe profit from deemphasizing technical details at this point and instead emphasizing the big crucial point: The diffeomorphism constraints are ‘solved’ without imposing Dirac constraint quantization.

I saw that Thomas Thiemann was in the audience, A. Ashtekar was, Hermann Nicolai was, and decided that it would be nice to continue our Coffee Table discussion about this point, which some people feel is a little problematic.

So as the talk was over I asked the lecturer about this point. He answered that this method is simpler than Dirac quantization and also has the advantge that also ‘large’ diffeomorphisms can be included, i.e. those that cannot continuously be connected to the identity (such as coordinate reflections).

When I began to argue that the method may be simpler but is not what Dirac tells us to do (for good reasons, like path integral and BRST formalism), A. Ashtekar approached me. He was very nice and helpful, as usual, and we went outside to further discuss things.

He checked if I am the one who had pointed him to the Coffee Table discussion and told me that he didn’t answer my email partly because he found some statements of the Coffee Table discussion overly offending. I think he is right about that and highly appreciate that he still very patiently talked and listened to me. Many thanks to Abhay Ashtekar, indeed.

First he said that the way LQG deals with the gauge constraints is not different from what one does in gauge theory. I replied that that has to do with the fact that the anomalies of the standard model happen to cancel, while it is not clear that those of canonical gravity would (without the ghost sector). I think he agreed.

I suggested that maybe LQG should then perhaps try to handle the ghost-extended Einstein-Hilbert action instead of the pure EH action. At least for 1+1 dimensional gravity this does indeed remove the anomaly, as is well known. I got the impression that A. Ashtekar found this idea is maybe worth considering (but I am not sure).

He told me that what he found problematic with the ‘LQG-string’ is that the method does not in any way seem to involve the fields on the background spacetime. I found this remark interesting, because it resonates with my own feelings concerning this point, which I have expressed here. A. Ashtekar hinted at some alternative approaches by himself and somebody else which are apparently under investigation, but I feel that I should not report on that here in public.

After this very illuminating discussion I was asked by somebody if I am working on LQG, because he had seen me pipe up in Fleischhack’s talk. After I had answered that to the negative we exchagned personal and scientific identities, and I was delighted to meet in Thorsten Prüstel somebody working on - guess what - discrete field theory on Lorentzian graphs.

We immediately had lots of things to talk about. I showed him Eric Forgy’s and mine pre-pre-print and he was very interested and invited me to give a talk about that at University of Hamburg. He himself is working on an interesting approach to get gravity from the nonunitary part of an extended gauge group on a graph field theory, roughly. I hope that when I am in Hamburg I will get the chance to learn more about that.

We already had to hurry to get to the ‘Welcome Party’. There we happened to sit next to Prof. Kastrup from Aachen. Kastrup was the thesis’ advisor of two of the leading figures in current LQG, namely Thomas Thiemann and Martin Bojowald. We learned how Thomas Thiemann as a student originally wanted to work on string theory and was later convinced to look into LQG.

I found it very interesting that Kastrup agreed with my assessment that LQG is not ‘canonical’ in the usual sense, because it does not represent the classical canonical coordinates and momenta as operators on a Hilbert space.

Indeed, in the afternoon I had heard the very interesting talk by Kastrup about quantization of integrable systems in angle/action variables. This is an interesting and subtle issue of canonical quantization, which can already be studied for the ordinary harmonic oscialltor.

The point is that by the naive correspondence rule one would think that, since the angle &omega; and the action S are a pair of canonically conjugate coordinates and momenta (like the ordinary q and p are, too) there should be self-adjoint operators &omega;^; and S^; which satisfy [&omega;^,S^]=i .

But it is easy to convince oneself that this cannot work, which has to do with the fact that the angle is defined only modulo 2&pi; , or, equivalently, that &omega; and S do not provide global coordinates on phase space, because the origin has the usual coordinate singularity of polar coordinates in the plane.

Kastrup showed how with taking much care one can instead construct two other operators K + and K - such that together with K 0 = S^ they give the Lie algebra of SO (1,2) and that from these the standard q^ and p^ can be reobtained. (This can be understood heuristically by thinking of the 2d phase space of the oscillator with the origin removed as a (‘light’-)cone.)

He concluded by saying that, while being equivalent to the quantization with q and p , this could have experimental consequences in quantum optics. (I asked him about how this can be true, but unfortunately failed to understand his answer.)

Anyway, this is a nice example for how subtle ordinary canonical quantization itself can already be.
------end quote--------
 
  • #12
Wow! I almost felt I was there. Thanks to you Marcus for posting it, and thanks to Urs for writing it!
 
  • #13
Did you guys know that audio and (in some cases) video of lectures can be downloaded at the ITP? For example:

http://online.kitp.ucsb.edu/cgi-bin/search-sched.pl?pwd=%2Fusr%2Fsrc%2Fslides%2Fonline&webpwd=http%3A%2F%2Fonline.kitp.ucsb.edu%2Fonline&file=schedule-index.html+*%2Fschedule-index.html&sidebar=pull%2Fsidebar.html&query=loop+quantum+gravity

http://online.kitp.ucsb.edu/cgi-bin/search-sched.pl?pwd=%2Fusr%2Fsrc%2Fslides%2Fonline&webpwd=http%3A%2F%2Fonline.kitp.ucsb.edu%2Fonline&file=schedule-index.html+*%2Fschedule-index.html&sidebar=pull%2Fsidebar.html&query=ashtekar

http://online.kitp.ucsb.edu/cgi-bin/search-sched.pl?pwd=%2Fusr%2Fsrc%2Fslides%2Fonline&webpwd=http%3A%2F%2Fonline.kitp.ucsb.edu%2Fonline&file=schedule-index.html+*%2Fschedule-index.html&sidebar=pull%2Fsidebar.html&query=baez
 
  • #14
Following the report on A. Ashtekar comments on the 'LQG-string', here is now the summary of an interview with H. Nicolai on LQG, anomalies and all that.
 
  • #15
Originally posted by Urs
Following the report on A. Ashtekar comments on the 'LQG-string', here is now the summary of an interview with H. Nicolai on LQG, anomalies and all that.

This is turning out to be a very interesting conference
(the German Physical Society Spring 2004) and includes even
the world premier performance of an opera about Albert Einstein!
I shall copy Urs CT post so that we have a complete account.

One may surmise that the role of Einstein is sung by the tenor and that the bass-baritone sings the part of Michele Besso, his friend for over 50 years. They may thus conveniently sing duets.



-----quote from Urs---
Tuesday
The talks today didn't interest me much (things like "Einstein and art") and I spent the time reading and answering my mail as well as preparing my own talk.

But I did have lots of very valuable conversations.

At lunch I met Hermann Nicolai and we had a long discussion about Pohlmeyer invariants, DDF invariants, LQG, the "LQG-string", diffeomorphism anomalies in string theory, string field theory, Matrix Models and prejudices in quantum gravity.

To begin with, I was kind of surprised to learn that H. Nicolai, together with K. Peeters, is currently thinking about Pohlmeyer invariants himself. We discussed the known results regarding their quantization and I mentioned that I think that there is a solution to the apparent problems. H. Nicolai was interested and invited me to visit the AEI in April to talk about these ideas.

Then of course we discussed the "LQG-string" and what can be learned from it about LQG itself. H. Nicolai said that he had hoped that the LQG people would find the anomaly for the string, which, as he said, he would have considered a breakthrough for the whole LQG field. But what has now actually been done, he said, reminded him more of certain artificial constructions in axiomatic field theory, which are also mathematically well defined but physically empty.

I asked him about what this now means for full LQG and he said that, similarly, he would expect that there should be an anomaly and that he finds the constructions done in LQG problematic.

I tried to understand how the same issue could be understood from within string theory, and he basically said that one would have to understand closed string field theory in order to tackle this question. But currently a satisfactory definition of closed string field theory is of course not known.

I said that I am wondering if we cannot learn anything in this regard from BFSS/IKKT Matrix Models. The authors of the IKKT/IIB model at least claimed that the permutation subgroup of the full U (N-> oo) gauge group of the model becomes the diffeomorphism group in the limit. Heuristically, we can think of the matrices as describing discrete spacetime points and the permutation subgroup permutes these points, so is related to diffeos in some sense.

Hermann Nicolai didn't know the details of this claim (me neither :-) and remarked that most permutations would give rather pathologic diffeos in the continuum limit. In any case, my uneducated guess is that the issue of diffeo anomalies and the like is hidden in the N -> oo limit of the matrix models, which is not well understood at all, as far as I know.

Since the time for the afternoon talk sessions was drawing near we stopped at this point. But as I was about to leave the cafeteria K.-H. Rehren approached me. I very much appreciated this, because from our previous online conversation I had gotten the, apparently wrong, impression that he wasn't interested in my comments on Pohlmeyer invariants.

I was delighted that we immediately sat down, pulled out pens and paper and began doing algebraic calculations. I think that in a couple of minutes we could clarify issues that would have taken weeks by email, at the previous speed.

But then we really had to hurry, because I was the one supposed to give the next talk!

I am not sure if it is a good or a bad sign, but at this DPG spring conference of the faculties "Gravitation and Theory of Relativity" and "Theoretical and Mathematical Foundations of Physics" my talk is apparently the only one directly concerned with strings! But I couldn't complain. With H. Nicolai, K.-H. Rehren and F. Mueller-Hoissen in the audience I knew I was talking to people who I would have liked to ask about their opinion on my stuff at any rate.

Unfortunately, there wasn't much time for questions and feedback. Let's see what tomorrow brings! If nothing else, my talk will probably enter the annals of the DPG as the only one based on chalk and blackboard in the age of PowerPoint. ;-)

After the afternoon sessions I took care to catch one of the speakers on NCG whose talks I had missed the day before. I was lucky to get hold of a mathematical physicist of the name Paschke, who had given a talk on NCG on pseudo-Riemmanina manifolds. He had presented a technique where you slice a globally hyperbolic manifold in compact spatial leaves, perform ordinary NCG a la Connes on these spatial slices and then figure out how to glue the resulting spectral triples together to get a spectral quadruple. At first it might seem that this way time is commutative while only space is noncommutative, but the crux is apparently that it turns out that this is not the case and that in some sense also time becomes noncommutative. But I didn't see this in any detail.

After having understood how Paschke is proposing to deal with NCG on pseudo-Riemannian spaces I tried to make him tell me what he thinks of Eric's and mine approach which is supposed to deal, among other things, with pseudo-Riemannian discrete spaces.

Paschke emphasized that he finds it very problematic to break the compactness assumption of Connes' approach, which technically means that one is (compact) or is not (non-compact) dealing with a C * algebra which has a unit, because then many of Connes' theorems won't hold if there is no unital C * algebra. That may be true, but I am under the strong impression that one can do interesting physics on non-compact noncommutative algebras nevertheless. But this will be a crucial point to be worked out if I want to communicate with the Connes school of NCG.

I have to run now to get to the debut performance (really, the official dress rehearsal) of Dirk D'ase's Einstein opera.
----end quote----
 
Last edited:
  • #16
the Josh/Urs conversation continues

today on SPR Josh replied to Urs:
----------quote--------

J:But it's also not necessary, because a self-adjoint representation of a Lie algebra exponentiates to a unitary representation of the Lie group. So what one should check instead is that one has an honest group representation, that is, that:

U(g_1)U(g_2) = U(g_1 * g_2)

And this *does* hold for the diffeo group in LQG.

U:Yes, as it does for Diff(S^1) x Diff(S^1) in Thomas Thiemann's 'LQG-sting'. By construction. I.e. operators U can be found which do satisfy this relation. I do understand and agree that these operators can be constructed.
The problem is, that these operators don't capture the usual quantum
effects.

J:I completely disagree. The point is that one could do *Schroedinger* quantum mechanics this way, if one wanted to; i.e., for the kinds of things we consider in the paper, one need never consider the self-adjoint generators. And you would get exactly the same physical effects as if you did consider them, because you're looking at exactly the same representation of exactly the same algebra. We are therefore just taking the unitary group rep, rather than the self-adjoint algebra rep, as the "jumping off" point.

U:They do not come from what ordinarily is called 'canonical quantization'. That's because the ordinary prescription of 'canonical quantization' at some point says that we are to promote the classical canonical coordinates and momenta to self-adjoint operators on some Hilbert space. This step is explicitly violated in LQG, since half of these objects are not represented as operators at all.
So with this step violated, something else has to be found to replace it. Among other things, I claim that this replacement introduces a large ambiguity.

J:Well, if you don't like ambiguity, you're going to find quantization
in general very frustrating :)

U:Namely whenever there is now an object in the classical theory in the enveloping algebra of p and q, you don't know precisely how to quantize it (even modulo operator ordering issues) since one of p and q is not represented on the Hilbert space.

J:Here is a good example of what I just said about ambiguity. You're
used to constructing the space of observables from the enveloping
algebra of the quantization of p and q. But of course this is hardly
ideal, from the point of view of "quantization," since the enveloping
algebra will not agree with the Poisson algebra of polynomials in p
and q, in general. So there is already ambiguity, in choosing what
subalgebra of the Poisson algebra one represents. People who study
the general mathematical problem of quantization worry a lot about
this sort of thing.

U:I think that this is a problem both in gr-qc/0207106 as well as in the 'LQG-string'. Here is why:
In the text leading to equation (IV.5) of gr-qc/0207106 you are looking for a way to express the notion of 'coherent state' in the LQG-like language, where you characterize a coherent state by its property of being an eigenstate of the annihilation operator
a ~ x + ip
(which is one of several possibilities of characterizing coherent states). This is an object in the algebra generated by p and q. It happens to have no operator ordering problems so the usual quantization of this guy is obvious.
But now in your framework p is not representable, so something else has to be done. Some object in the algebra of operators that you do have must be identified as being the 'quantization' of the annihilation operator. Due to the Weyl-algebra context, what you really want is the exponentiation
of the adjoint of the annihilation operator, i.e. an analog of

exp(a^dag) .

This has to be expressed somehow using the operators x and V that do exist on your Hilbert space. You do this by looking at the usual quantization of this object and read off the expression above equation (IV.5) from this, e.g.

exp(a^dag) -> exp(x)V(-1/2)e^(-1/4) .

But why don't you use for instance
exp(a^dag) -> exp(x)V(-1/2)
or
exp(a^dag) -> exp(x)V(-1/2)e^(42) ?

This is the factor which I am talking about and which is put in by hand. It is _not_ put in by hand in the _usual_ quantization, because there the e^(-1/4) comes from the BCH theorem. But in your rep there is no p and q which could enter the exponent in the BCH theorem. You are just copying this plausible looking result to your representation.

J:You are misunderstanding several things here. First, as I indicated in my original post, this is not the point in the paper where we define the physics. That has already been done, in choosing the algebra and representation; the rest of the paper (more or less) is about finding out what the physics of this rep is---not specifiying it.
Even though there is no p operator (there is in fact a q operator) we
still know what the commutation relations should be, and therefore how
to apply the CBH theorem, even if we had never heard of the
Schroedinger rep. That's because we know the what the commutation
relations should be from the *classical* theory, the Poisson algebra.
This alone tells us that

exp(a^dag) -> exp(x)V(-1/2)e^(42)

is wrong, provided we know that a^dag = q +ip classically. But how do
we know that? In other words, what does a state being an eigenstate
of this particular operator have to do with that same state being a
semiclassical state?

But we could just as well ask the same question in Schroedinger
quantum mechanics! The point is that as far as semiclassicality is
concerned (because of course as you know coherent states have many
other interesting and important properties), it's just a fact that the
states which satisfy this eigenvalue equation also have minimal
uncertainty and saturate the Heisenberg bound. But it is the latter
that makes them semiclassical, not the fact that they are eigenstates
of the annihilation operator. Likewise, in our paper the states are
semiclassical because they meet the criteria on pp. 20-21. It just
turns out that states which meet this criteria also satisfy the
eigenvalue equation (IV.5).


U:And from reading the paper I got the impression that you would agree that here you are fixing an ambiguity by hand. Because on p.14 it says:
"Collecting these ideas motivated by results in the Schroedinger
representation, we are now led to seek the analog...".

J:You're misunderstanding our point, I think.
We've introduced in this paper a non-standard rep of the
Weyl-Heisenberg algebra, which we claim should be thought of as a
viable (at least for a non-relativistic theory) quantization of the
particle on the line. If we wanted to do the analagous thing to
quantum gravity, all we would be looking for is the semiclassical
limit: that is, to show that Newton's laws are recovered in an
appropriate sector of the theory. Of course, for quantum mechanics
(as opposed to quantum gravity, at present) this is not nearly enough;
we must also show that the quantum effects that are confirmed by
experiment are also reproduced.

But in fact we aim to do slightly more even than that. We are trying
to show that a great deal of the structure of the Schroedinger
representation can be carried over to this polymer particle
representation. A priori that might seem impossible, since the two
Hilbert spaces aren't even unitarily equivalent, much less the
representations. The "conventional wisdom" would therefore dictate
that these must be physically different theories. And indeed they
are: there is some regime in which their physical predictions will
disagree, as there must be. But not only is there a large regime in
which their predictions are physically indistiguishable, but there is
much more similarity in structure than that superficial observation of
unitary inequivalence might have suggested. And in order to compare
the structure of the polymer particle rep to the Schrodinger rep, we
obviously have to say what the latter is. But if we didn't know or
care about the Schrodinger rep, there would be no need to rely on it.

But we do, and that is why the emphasis on constructing an analogue of
the coherent state eigenvalue equation; if all we cared about was
semiclassicality we could have gone directly to positing the state and
verifying its properties. This might have been inelegant, but there's
nothing per se that requires an eigenvalue equation to select
semiclassical states. And indeed most semiclassical states will not
satisfy any particular such equation, because the class of
semiclassical states is much larger than the class of coherent states.
But in fact these states, which are exactly the states one might have
"guessed," also satisfy the same criterion as coherent states in the
Schrodinger representation.

-----end quote----
 
  • #17
a few more bits of dialog

-----quote from today's SPR post----
U:I do agree that this is indeed the most natural choice for this simple system. My problem is that in more difficult cases, where there is mayber no working Schroedinger representation from which ideas can be motivated, how do you fix such ambiguities then?

J:Two things: again, I don't really think there's an "ambiguity," as I explained above. Second and more importantly, there are guidelines to what one should be looking for in other quantum theories, including gravity. Specifically, one would be looking for the "best approximation" to Fock space, among other things. If you're really interested in this then you should read the papers that led up to the one we wrote, namely:

gr-qc/0001050, gr-qc/0104051, and gr-qc/0107043

Some other stuff has been put on the web since then as well.

U:I noted that in the 'LQG-string' as well as for the diffeo-constraints of 3+1d LQG gravity the analog of this ambiguity is fixed by using ideas motivated by results not of the Schroedinger rep but of the classical Poisson algabra.

J:As I said ealier, you're comparing apples to oranges if you compare the selection of coherent states in the polymer particle paper to the representation of the diff-algebra in LQG. The former isn't representing any algebra at all; the choice of algebra has already been made, and it was made based on the classical Poisson algebra. In this respect the polymer particle is not a complete model because we didn't look at constraints, but if we had they too would be dictated by the Poisson algebra; you would fix that before you ever looked for semiclassical states, coherent or otherwise.

--Josh
--------------end quote---------
 
Last edited:
  • #18
I am now, thoroughly confused again. Proffessors I've talked too are as well it seems (eg the answer depends on who you ask).

One thing is for sure, the LQG approach to this particular problem is subtle in many ways. In many ways its conjecturing correspondances between classical regimes and quantum ones, and inserting them by hand.

I don't buy really buy it, but again the mathematical formalism deep at the heart of this is abstracted away from the language they use.

Back to the blackboard!
 
  • #19
Originally posted by Haelfix
I don't buy...it

Hardly anyone does.
 
  • #20
the Josh/Urs dialog continued on SPR yesterday

----from yesterday's post on spr---
JOSH: In the case of the diffeo group in LQG, one can explicitly check that the Hilbert space used, which carries an appropriated rep of the Weyl-Heisenberg algebra and other observables, like the area operator, does in fact have a unitary rep of the diffeo group as well.


URS: This is apparenly a crucial point for the communication between LQG people and others. In LQG there is a lot of emphasis on finding such unitary reps. For instance Thomas Thiemann shows how such a rep exists for the string. But I believe that it is important to emphasize that the mere existence of such reps is not the issue. They do exist, and it is not hard to see why and how.


JOSH: I don't think this is nearly as easy as you claim. The key point is finding all of the things I mention in the quoted para. in one rep.

To back up a little, it's best to outline some of the criteria normally needed for a "quantization" of a given classical system. Normally, for a prequantization one needs:

(1) Poisson brackets go over to commutators,
{f,g} --> (i/\hbar)[\hat{f},\hat{g}]
(2) The constant function on phase space is mapped to the identity operator.
(3) If the Hamiltonian vector field of f on phase space is complete, then \hat{f} is essentially self-adjoint.

If we have only these three postulates, then there are "quantizations" such that the entire Poisson algebra can be promoted to an operator representation preserving the classical Poisson brackets.

However, even for finite dimensional systems this is not enough, and so for a prequantization to be a true quantization, one demands another criteria, usually either that the representation of the "basic observables" (i.e., the p's and q's) be represented irreducibly, or that for instance a certain class of observables (functions only of q, for instance) be multiplicatively represented. This severely constrains allowed quantizations, and in general one can then only implement (1) for a certain subalgebra of the classical symplectic algebra.

Now consider the case where one is quantizing a constrained system by the Dirac quantization "program." Then it is essential that the constraints be among the classical observables that can be promoted to operators respecting the Poisson algebra, and in general this is nontrivial.

So far, everything I have written is for what I think you would call "canonical quantization." What we are doing is now demanding instead of (1) and (3) that a subgroup of the symplectic group on the manifold---corresponding to the subalgebra that is represented by e.s.a. operators---be unitarily represented on the Hilbert space. Appropriate irreducibility criteria must still hold. When the unitary representation is continuous, one can obtain a representation of the self-adjoint operators from the given unitary representation; when it is not, no such rep exists, and we are now allowing for this possiblity

Now, it seems that you think that this relaxation makes it easy to satisfy the criteria I list above. Why do you think this? Can you give a general construction? I'll tell you why I don't think it's so easy, again using LQG as an example.

The way that these criteria are met in LQG is as follows. One takes as the basic algebra of observables the "flux-holonomy algebra"; this is supposed to be analagous to the Heisenberg algebra in ordinary QM. One then demands that the holonomy observables be implemented multiplicatively. This is done by constructing a certain space Abar of "generalized connections" on which the holonomy algebra is the algebra of all bounded, continuous functions. Then the Hilbert space you have is an L_2 space on Abar with respect to some measure, and the choice of representation is dictated by the choice of this measure. *Any* regular, borel measure by construction will represent the holonomy algebra faithfully. But that's not enough. We must represent the entire flux-holonmy algebra, and to do Dirac quantization we must also represent the diffeomorphism group unitarily.


Take the latter restriction first. This means we can no longer consider any measure on Abar, but just diffeomorphism invariant measures. That already throws out a lot of possibilities.


However, there are still loads of diff-invariant measures on Abar. But there is only *one* for which it is true that one gets a representation of the full flux-holonomy algebra. This is the measure usually referred to as the Ashtekar-Isham-Lewandowski measure (or sometimes just Ashtekar-Lewandowski). That this is the unique measure Abar allowing one to represent both the flux holonomy algebra and the diffeomorphism algebra is shown in recent work of Sahlmann, Thiemann, Lewandowski, and Okolow.


So, given this I am highly dubious that one can make a general construction for any subalgebra of the Poisson algebra; it is not, for instance, even clear that one can enlarge the flux-holonomy + diffeomorphism algebra in LQG without breaking the construction.


URS: The crucial problem is rather that many people don't want to call the procedure of finding such a unitary rep a 'quantization' of the system at hand. Rather, they would want to see a canonical quantization in the ordinary strict sense where the symmetry generators are represented on some Hilbert space.

JOSH: I suppose some people may think that, but it's worth pointing out that people other than the LQG community (for instance many working in various rigorous approaches to QFT) often focus on unitary reps, for several reasons.

(1) Unitary operators, being bounded, are much nicer to deal with than unbounded self-adjoint operators.

(2) The unitary criteria is easier to justify physically, since it is what corresponds to preservation of expectation values by the action of a symmetry. When the rep is continuous and the phase space finite-dimensional, the Stone-von Neumann theorem guarantees that the only continuous unitary rep is unitarily equivalent to the Schroedinger rep. But in infinite dimensions the Stone-von Neumann theorem fails, and it is for this reason that in QFT one often wants to look at the unitary algebra. Haag mentions this early on in _Local_Quantum_Physics_, IIRC.

(3) When the symmetry group is infinite dimensional, as for instance Dif(S^1), any neighborhood of the identity contains group elements that cannot be obtained by exponentiating the Lie algebra. Thus, the algebra doesn't carry the full information about the group.


JOSH: But there's no claim that for any classical system
with a Lie algebra of symmetries, there must exist some quantum rep in
which the corresponding group action is unitarily implemented.


URS: On the other hand, it seems that when you allow non-seperable Hilbert spaces then constructions analogous to those used in the 'LQG-string' do give you unitary reps for virtually everything. Can you give an example of a symmetry group which, by the methods used in LQG and in the 'LQG-string' could _not_ be represented unitarily?


JOSH: Again, I would instead challenge you to give examples of how you can do this for "virtually everything."


JOSH: Even if there is such a rep, there can of course be other pathologies of the quantization.


URS: Could you please explain what kind of pathologies you have in mind here?


JOSH: Failing one or more of the criteria I mentioned above, for starters.


URS: For instance, do you think that the method of 'shadow states' could be applied to Thomas Thiemann's 'LQG-string', thus showing that his approach does reproduce the usual quantization in some limit?


JOSH: I don't know---we haven't even tackled full GR yet in this framework. And I haven't looked at Thomas's paper.


URS: I believe that this won't be possible, because all the information about the usual quantum effects have been eliminated in the 'LQG-string' and they won't reappear in any limit.


JOSH: I don't understand this statement at all. See my other response (that I hope to write :) ) to your other reply to my post. But at various points you seem to me to have implied that the absence of anomalies means that one won't get the "usual quantum effects." But there are lots of quantum effects besides the existence of anomalies! After all, plenty of systems don't show anomalies in standard quantum prescriptions. Surely you wouldn't say that these show no quantum effects, that the classical and quantum systems are physically indistinguishable?


For instance, one of the first quantum effects one learns about is the inability to measure momentum and position simultaneously, because they are not commuting observables. This in turn arises because one is quantizing using the Poisson brackets, rather than the usual commutative algebra structure of real-valued functions on phase space. And this effect is certainly preserved in these "unitary group" quantizations, precisely because (for a particle on the line, for instance) one chooses to represent the Heisenberg group, rather than the commutative group R^2.


URS: Is this the kind of pathology that you are referring to above?


JOSH: No.

----endquote---
 
  • #21
more of yesterday's Josh and Urs dialog

---from yesterday's SPR discussion---
JOSH: I completely disagree. The point is that one could do *Schroedinger* quantum mechanics this way, if one wanted to; i.e., for the kinds of things we consider in the paper, one need never consider the self-adjoint generators. And you would get exactly the same physical effects as if you did consider them, because you're looking at exactly the same representation of exactly the same algebra. We are therefore just taking the unitary group rep, rather than the self-adjoint algebra rep, as the "jumping off" point.


URS: I don't doubt that you can choose the U in such a way that the usual quantum effects are respected. But they are not always chosen that way. The U(phi) in Thomas Thiemann's paper are chosen in such a way that the usual quantum effects (the anomaly) are absent. He could have chosen U(phi) which reproduce the anomaly. But apparently in the LQG-framwork there is the freedom to make different choices. This is the ambiguity that I mentioned.

I stated that I think that obtaining or not obtaining the anomaly in Thomas Thiemann's 'LQG-string' is completely analogous to including or not including that factor exp(-alpha^2/2) in your paper. Since you disagree it would be helpful if we could reverse the burden of proof:

Do you think, and if you do could you please show, that the technique of 'shadow states' which you have tested on the 1d nonrelativistic particle, can be applied to Thomas Thiemann's 'LQG-string' such that the results of the usual quantization are reproduced in an appropriate limit?

This would be very helpful.

JOSH: Well, if you don't like ambiguity, you're going to find quantization in general very frustrating :)

URS:Right. Because I knew that I was emphasizing the ambiguities of first quantization from the very beginning of the discussion
(http://golem.ph.utexas.edu/string/archives/000299.html). But the usual operator ambiguities are something different from the ambiguities that are introduced in the LQG framework. And I was trying to stress that in my last
post. Neither the equation

a |psi> = alpha|psi>

characterizing coherent states nor the Virasoro anomaly

[L_m,L_n] = (m-n)L_m+n + (D/12)m^3 + O(m)

are sensitive to operator ordering issues. So there is no ambiguity here in the usual canonical quantization prescription. But as soon as you are proposing a quantization where neither the operator "a" nor the operators "L_m" even exist the above are no longer universal facts. Thomas Thiemann claims that he may use operators U(phi) analogous to (but not related to) exponentiated generators L_m such that there is no anomaly. In the same vein one could choose the usual exp(L_m) such that the anomaly is there. So in such a context there is apparently a much larger ambiguity.

JOSH: Even though there is no p operator (there is in fact a q operator) we still know what the commutation relations should be, and therefore how to apply the CBH theorem, even if we had never heard of the Schroedinger rep.

If this were true, then why don't you apply this reasoning to the spatial diffeo generators D_i(x) in LQG, or why didn't Thomas Thiemann apply this reasoning to the 'LQG-string', thereby rediscovering the anomaly?

If you can guess the analogue of the BCH theorem even in the absense of a Schroedinger rep (namely no usual consistent quantization of the ADM constraints D_i(x) is known) then why isn't that discussed and applied in LQG? What is the CBH analogous correction to
exp({int f(x) D_i(x) , .})exp({int f(y) D_j(y) , .})?



JOSH: As I said ealier, you're comparing apples to oranges if you compare the selection of coherent states in the polymer particle paper to the representation of the diff-algebra in LQG. The former isn't representing any algebra at all; the choice of algebra has already been made, and it was made based on the classical Poisson algebra.

URS: The analogy that I am seeing is that in both cases you have an operator equation which characterizes certain states, namely coherent states in one case and physical states in the other case. Now the operators entering these operator equations are not available in your framework and so something else is sought for.

But if you don't agree that the crucial non-standard step of the
'LQG-string' and of LGQ of 3+1d gravity is also present in the 'shadow
states' paper (although dealt with differently in both cases) then I'd
consider it more productive for our discussion if we could reverse the way of proof: Instead of me arguing why the 'shadow states' technique won't work for Thiemann's 'LQG-string' you could demonstrate that and how it does. For instance it would be great if you could show that, by taking some limit if necessary, you can find the tachyon 'shadow state' of the 'LQG-string'.

But having said that, I recall that probably the absence of this state will be considered an advantage. This then would leave me completely puzzled: Why is it ok for the 1d nonrelativistic particle that the LQG-like quantization approximates the usual quantization, while for the Nambu-Goto action it is not?

---endquote---
 
Last edited:
  • #22
This 'copy and paste' is a fun game! :-)

I can even provide you with spr messages that haven't even appeared yet ! ;-)

------------

"Josh Willis" <jwillis@gravity.psu.edu> schrieb I am Newsbeitrag
news:c3goq8$18le$1@f04n12.cac.psu.edu...

> (1) Poisson brackets go over to commutators,
> {f,g} --> (i/\hbar)[\hat{f},\hat{g}]
> (2) The constant function on phase space is mapped to the identity
operator.
> (3) If the Hamiltonian vector field of f on phase space is complete,
> then \hat{f} is essentially self-adjoint.
[...]
> So far, everything I have written is for what I think you would call
> "canonical quantization."

Yes.

> What we are doing is now demanding instead of
> (1) and (3)

Just for the record: You are now saying that what is done in LQG ('we') is
not canonical quantization as sketched above but a modification
('relaxation') thereof.

> that a subgroup of the symplectic group on the
> manifold---corresponding to the subalgebra that is represented by e.s.a.
> operators---be unitarily represented on the Hilbert space. Appropriate
> irreducibility criteria must still hold. When the unitary
> representation is continuous, one can obtain a representation of the
> self-adjoint operators from the given unitary representation; when it is
> not, no such rep exists, and we are now allowing for this possiblity

Agreed and understood. The point is that it is not clear that allowing for
this possibility is physically viable.

> Now, it seems that you think that this relaxation makes it easy to
> satisfy the criteria I list above. Why do you think this?

I think that if you could satisfy the above criteria without the relaxation
for 1+3d gravity then you would do that instead of proposing a speculative
relaxation of these criteria. Not? Isn't the problem with doing a ("what I
would call") strict canonical quantization of gravity, along the unrelaxed
lines sketched above, precisely the reaon why these relaxations are
considered in the first place?

> Can you give a general construction?

See below.

> > The crucial problem is rather that many people don't want to call the
> > procedure of finding such a unitary rep a 'quantization' of the system
> > at hand. Rather, they would want to see a canonical quantization in
> > the ordinary strict sense where the symmetry generators are
> > represented on some Hilbert space.
>
> I suppose some people may think that, but it's worth pointing out that
> people other than the LQG community (for instance many working in
> various rigorous approaches to QFT) often focus on unitary reps, for
> several reasons.

Ok, yes. It appears that this is the main result of all the discussion
initiated by the 'LQG-string':

LQG uses methods that many people would not call quantization while some
people (most notably in the fields of LQG and AQFT) would.

For me this is already a very useful information. Before this whole
discussion I was told and did believe that LQG is a very conservative
approach towards quantizing gravity. I had always thought "Well, just doing
canonical quantization of something cannot be that wrong."

Now I learned that LQG is 'canonical' only in a generalized sense which is
not generally accepted and which furthermore produces strange results when
applied to systems that we actually do understand (as opposed to full
nonperturbative quantum gravity).

> Again, I would instead challenge you to give examples of how you can do
> this for "virtually everything."

What I have in mind is that the way exact reps of the classical groups are
constructed for the 'LQG-string' as well as for LQG of 3+1d gravity follows
the following presciption:

- Given any (infinite dimensional) group G with elements phi.

- Pick any set of objects psi(s) such that the group acts faithfully on
these like, say,
phi[psi(s)] = psi(phi(s)) .

- Construct a Hilbert space such that the phi(s) are orthonormal elements

<phi(s),phi(s')> = 1 iff s=s' and = 0 otherwise .

- On this Hilbert space the classical group is represented _exactly_ (not
projectively or something) by the operators U(phi) defined by

U(phi)|psi(s)> := |psi(phi(s))> .

Wether or not the U(phi) are unitary is a different question. The important
point is that this way the classical symmetry group can always be
transferred to the 'quantum theory' exactly, by construction.

This is the essence of what Thomas Thiemann does in the 'LQG-string' paper.
And I think it is also what is done for the spatial diffeomorphisms in LQG
of 3+1d gravity.

> And I haven't looked at Thomas's paper.

Oh, too bad! :-) Thomas Thiemann's paper has the virtue that it applies LQG
methods to a system which is not quite as trivial as the free 1d nonrel
particle, but still much better understood than full quantum gravity. It
seems to be an ideal testing ground for LQG methods.

> > I believe that this won't be possible, because all the information
> > about the usual quantum effects have been eliminated in the
> > 'LQG-string' and they won't reappear in any limit.
>
> I don't understand this statement at all. See my other response (that I
> hope to write :) ) to your other reply to my post. But at various
> points you seem to me to have implied that the absence of anomalies
> means that one won't get the "usual quantum effects."

This stratement is a little vague, true. But I think from the context it
ishould be quite clear what I mean. Both the corrections that come from
applying the BCH theorem to exp(p)exp(q) as well as the anomaly in the
Virasoro algebra come from the noncommutativity of the quantum algebra which
is absent in the classical algebra. These effects cannot be found if the
respective algebra is not present. So in the 'LQG-string' they are simply
absent while in your 'shadow states' paper you argue that they must be taken
into account.

> But there are
> lots of quantum effects besides the existence of anomalies! After all,
> plenty of systems don't show anomalies in standard quantum
> prescriptions. Surely you wouldn't say that these show no quantum
> effects, that the classical and quantum systems are physically
> indistinguishable?

No, of course not. It just happens that for the perhaps most intersting
testing ground example of LQG methods, namely the 'LQG string', the anomaly
is at the center of attention. The whole point for Hermann Nicolai to
propose that LQG methods should be applied to the string was to see if the
anomaly can be found this way. He told me that had considered it a
"breakthrough for the field of LQG" if it would have been found.
 
  • #23
Hi to Urs

Hello Urs, thanks for helping bring your interesting discussion to PhysicsForums!
In reading this BTW it did seem to me that you are overly focussed on Thomas Thiemann's "Loop-String" paper.

This paper does not seem to have received much attention among LQG researchers. Indeed
(1) Josh Willis mentioned yesterday he had not even looked at it
(2) Stingray said in another PF post that he had asked several
loop quantum gravitists about it---but not much response
(3) soon after TT posted the paper he attended a conference
that Corichi reported on the web. there was no report of any discussion of "Loop-String" although Thiemann talked on other subjects.

You see, I get the impression that the "Loop-String" paper made a much bigger impression on String researchers and got a lot more discussion with them than it did with LQG folks. It even seems kind of distracting to want to keep going back to it, when you are discussing with Josh Willis the paper he wrote with Ashtekar. But this is merely my personal opinion. No doubt you have some good reason for viewing TT paper as very important and wanting to get JW comments on it.

I must say that your reportage ("first person journalism") from Ulm was great! exciting and refreshing to get a lively first-hand report from such a dynamite conference! Also it impressed me that you had such a long private discussion with Ashtekar.
 
Last edited:
  • #24
Marcus -

it may well be that the 'LQG-string' doesn't receive much attention among LQGists. But I am getting the impression that the problems that this paper highlighted should receive more attention.

In my humble opinion the free 1d nonrel particle discussed by Willis et al. is so trivial that it can fool you into not realizing that there are problematic steps. The 'LQG-string' is much better suited in that respect. The NG action is very well understood but not as trivial as the free particle.

But you are right, I don't know how much longer I will have the energy arguing that there are problematic steps if the people working on LQG won't agree with this assessment. Well, my goal is not to fix what I (and many others who explicitly told me so) consider problems of the LQG approach.
 
  • #25
Yes, the Stingray comment was in the "Research demographics again"
thread

https://www.physicsforums.com/showthread.php?s=&postid=167993#post167993

Someone had posted a fantasy that the Coffee Table criticism of TT's paper was so deep and fundamental for LQG that the quantum gravitists would now be puzzling about this and so he just remarked that this seems not to be the case

--------exerpt from Stingray post----

quote:
------------------------------------------
As a result, the very small number of people currently doing LQG are scratching their heads over this.
------------------------------------------

From talking to people working in LQG, this is not correct. There were certainly issues with that paper, and I wish I remember what those were right now. They were not fundamentally damaging to the ideas of LQG. I highly doubt that all string papers are perfect either...

---------end quote-----

this is one of the reasons I feel you are over-focussed on that
not-very-representative paper and I would hope that you could turn your attention to a very central recent paper like the one by Fairbairn and Rovelli. I am certain you would find there MUCH to displease you :wink: The paper is a leadingedge high-risk venture, with significant details to be worked out, it seems to me.
the separability of the kinematical Hilbert space
 
Last edited:
  • #26
Originally posted by Urs

...I don't know how much longer I will have the energy arguing that there are problematic steps if the people working on LQG won't agree with this assessment. Well, my goal is not to fix what I (and many others who explicitly told me so) consider problems of the LQG approach.

I sympathize strongly because I understand from Eric F that you are in graduate school where one's own research is almost a survival issue!

I would certainly understand you focussing your obvious talent and energy on your own favorite topics and not on "fixing the problems of LQG!"

If I were to try to sum up, in the simplest possible way, how you see these problems which you mention, I suppose I would say this (Is this correct?)

I believe you think that diffeomorphism invariance should be implemented by operator equations on the kinematical state space----constraint equations. these operators will correspond to generators of the diffeomorphism group.

I believe you wouold reject the fairly elementary and transparent construction on page 4 of
http://arxiv.org/gr-qc/0403047 [Broken]
(the Fairbairn/Rovelli paper)
which defines a projection map from a preliminary hilbertspace
containing some unphysical redundancy down to the more concrete kinematical state space. This seems to be the thrust of your comment,
or perhaps you would describe it differently.

I am still trying to apply your reasoning to Rovelli's book or to this paper and to see clearly whether it amounts to anything *in that context*. I want to see what, if anything, your comments say about LQG proper not just about the Thiemann "Loop-String" paper, which (although it got quite some attention from those doing string research) seems somewhat periferal to the main body.
 
Last edited by a moderator:
  • #27
the conversation continued today

Thomas Larsson replied to Josh Willis, adding another voice to
the Josh/Urs conversation:

----quote from today's post by Larsson---

Josh Willis <jwillis@gravity.psu.edu> wrote:

> (3) When the symmetry group is infinite dimensional, as for instance
> Dif(S^1), any neighborhood of the identity contains group elements that
> cannot be obtained by exponentiating the Lie algebra. Thus, the algebra
> doesn't carry the full information about the group.
>


There are also exponented vector fields that are not
diffeomorphisms. My impression is that this distinction is a
mathematical subtlety without real physical significance. The
important thing is that the diffeomorphism group and the algebra
of vector fields share essentially the same representations; the
classical irreps are tensor densities and exact forms.


Most interesting Lie algebra cocycles can be integrated to group
cocycles, e.g., the Virasoro algebra gives the Schwarzian
derivative. Yuly Billig has recently written down the
analogous group cocycles in higher dimensions.


> > I believe that this won't be possible, because all the information
> > about the usual quantum effects have been eliminated in the
> > 'LQG-string' and they won't reappear in any limit.
>
> I don't understand this statement at all. See my other response (that I
> hope to write :) ) to your other reply to my post. But at various
> points you seem to me to have implied that the absence of anomalies
> means that one won't get the "usual quantum effects." But there are
> lots of quantum effects besides the existence of anomalies!


Urs Schreiber's arguments have some interesting logical
consequences. If canonical quantization of the string leads to
conformal anomalies (only as an intermediate step before
introducing ghosts, but anyway), canonical quantization of
gravity should lead to similar intermediate diffeo anomalies.
However, pure gravitational anomalies in field and string theory
can only arise in 4k+2 spacetime dimensions. Thus string theory
is also incapable of producing such diffeo anomalies in 3D and
4D.


Note the difference compared to anomaly cancellation: here no
anomalies can be written down at all. The reason is that the
relevant anomalies are functionals of the observer's trajectory,
which is simply not available unless you introduce it.


It thus seems that Urs Schreiber's argument, if correct, proves
that string theory is wrong.
-----------end quote----------
 
Last edited:
  • #28
I must say that your reportage ("first person journalism") from Ulm was great!

Thanks. I have now also included a few photographs. See here (you have to scroll down a little).
 
  • #29
Originally posted by Urs
Thanks. I have now also included a few photographs. See here (you have to scroll down a little).

Whoah! The statues of Albert sticking his tongue out are terrible kitsch!

I like the stark machine-look of the Ulm university architecture, glad you included that snapshot of the building.

But who is the smiling guy who has intruded into your lecture
about field theory? He is getting in the way of the overhead projector.

If you keep on putting snaps of people like Ashtekar on line you will make Coffee Table famous and we will have to pay you an honorarium to get you to come here to PF.

thanks for the link in spite of it all
 

What is the diffeomorphism group and why is it important in LQG?

The diffeomorphism group is a mathematical concept that represents the set of all possible transformations that can be applied to a space without changing its fundamental properties. In Loop Quantum Gravity (LQG), the diffeomorphism group is important because it allows for the understanding of how space-time can be described as a discrete network of interconnected loops, rather than a continuous fabric.

What are the main differences between LQG and other theories of quantum gravity?

One of the main differences between LQG and other theories of quantum gravity, such as string theory, is the approach to understanding the fundamental nature of space-time. LQG focuses on the discrete nature of space-time, while string theory proposes a continuous space-time fabric. Additionally, LQG does not require the existence of extra dimensions, unlike string theory.

How does LQG incorporate the principles of general relativity?

LQG incorporates the principles of general relativity by using the concept of space-time curvature to explain the dynamics of gravity. In LQG, space-time is described as a network of interconnected loops, and the curvature of this network is used to describe the geometry of space-time and the effects of gravity.

What are some potential implications of LQG for our understanding of the universe?

LQG has the potential to provide a deeper understanding of the fundamental nature of space-time and gravity, which could have implications for our understanding of the universe at both the macroscopic and microscopic levels. It may also provide a way to reconcile general relativity and quantum mechanics, two theories that currently have not been unified.

What are some current challenges facing LQG and its development as a theory?

One of the main challenges facing LQG is the difficulty in connecting it with other theories, such as quantum field theory, which is necessary for a complete understanding of the universe. Additionally, LQG is a relatively new theory and is still undergoing significant development and refinement, which can be a slow and complex process. Finally, there are also challenges in testing and verifying the predictions of LQG through experiments and observations.

Similar threads

  • Beyond the Standard Models
Replies
13
Views
1K
  • Beyond the Standard Models
Replies
9
Views
3K
  • Beyond the Standard Models
Replies
5
Views
2K
  • Beyond the Standard Models
Replies
24
Views
4K
Replies
12
Views
2K
  • Beyond the Standard Models
Replies
2
Views
2K
  • Beyond the Standard Models
Replies
2
Views
2K
  • Beyond the Standard Models
Replies
2
Views
1K
  • Beyond the Standard Models
Replies
9
Views
1K
  • Beyond the Standard Models
Replies
7
Views
2K
Back
Top